Superdeterminism says that we can observe violations from Bell's constraint for local hidden variable theories, if we assume the measurement choices are correlated, which they must be in a deterministic theory where all measurement choices are pre-decided at the Big Bang. It is correct that Bell's constraint gets violated if we assume correlations between measurement choices. But this does not automatically mean that the new hidden variable theory predicts the correlations exactly as predicted by quantum mechanics.
If we consider that all measurement choices are correlated since the Big Bang, then those correlations would effectively be randomised. So, wouldn't it be impossible for the hidden variable theory to predict the exact correlations as predicted by Quantum Mechanics?
This is because Quantum Mechanics only considers the Bell test in isolation, in order to predict the experimentally observed correlations. If two experimenters do two different Bells experiments, Quantum Mechanics predicts the same correlation both times. But a superdeterministic theory's predictions would depend not just on the experiment, but on everything that happened since the Big Bang.